IT surveys usually do not provide surprising results. They often confirm what we already suspect to be true.
But a recent global study from IDC, which SAS commissioned, did offer some counterintuitive findings. They are based on two measurements:
- Trustworthy AI Index: This tracks how enterprises invest in responsible, reliable and ethical AI practices.
- Impact Index: This measures the real value realized from AI. This includes productivity, innovation, customer experience, and financial returns.
What the trust dilemma means is that enterprises are either underusing reliable systems because they lack confidence or relying on unproven tools that haven’t demonstrated reliability. The result is disappointing ROI and heightened risks for deployments.
Bryan Harris, CTO at SAS, warned against giving AI systems “trust just because they’re sophisticated.” Instead, he argues, they must earn it. This must be done by demonstrating integrity, transparency, and reliability across what he calls the “decision supply chain.” Each AI-driven decision depends on upstream choices about data quality, precision, and human oversight.
Impact of Generative AI
There’s one part of the IDC survey that is in line with the conventional wisdom: 81% of enterprises use generative AI compared to 66% that use predictive models.
Yet there is something that is counterintuitive. The survey found that respondents trust GenAI 200% more than machine learning, regardless of the errors and hallucinations.
Why this result? It seems that there is a “humanlike bias.” This is where we trust what is familiar and intuitive, not necessarily what is accurate.
The Agentic AI Future
Of course, today’s buzzy category is agentic AI. This is about multiple systems that work collaboratively and autonomously to carry out tasks.
The IDC survey shows that 52% of organizations have adopted some form of agentic AI. But success is far from easy, requiring a strong data foundation, solid governance, and specialized skill sets for employees. Keep in mind that almost half of the respondents in the survey say they fall short on these areas.
Agentic AI is not a technology that should be broadly applied. The fact is that it can be expensive, in terms of token consumption, as well as suffer from high latency and challenges with selecting the right tools. In fact, for many scenarios, machine learning or even traditional rule-based automation may be a better approach.
“You have to analyze what you want the end decision to be, back it up into the other decisions that support it, and then apply the right AI or rules to service that decision supply chain,” said Harris.
The Path Forward
IDC’s findings make one thing clear: trust isn’t just a moral imperative—it’s a financial one. Organizations that invest in governance, transparency, and responsible AI frameworks are 60% more likely to double ROI from their AI projects.
In other words, the next wave of AI progress won’t come from sheer scale or speed, but from earning trust through structure, discipline, and design.
